CN103269295A - Path capacity measuring method based on packet pair - Google Patents

Abstract

The invention provides a path capacity measuring method based on a packet pair. When the up link path capacity is measured, the detection packet pair is transmitted from a source site to a destination site, the destination site processes arriving and spreading time of the detection packet pair, and the up link path capacity is further calculated. When the down link path capacity is measured, a detection packet can be transmitted from the destination site to the source site, and the same mode is adopted to calculate the down link path capacity. The detection packet transmitted from the destination site to the source site can be an ACK response packet, the detection packet contains up link path capacity information, and therefore the up link path capacity and the down link path capacity can be measured at a time.

Description

Method of measurement based on the right path capacity of bag

Technical field

The invention belongs to network communication field, particularly a kind of method of measurement based on the right up-downgoing path capacity of bag.

Background technology

Path capacity, weighing bottle neck bandwidth (Bottleneck Bandwidth) again, the basic index as the largest data transfer ability on the network end-to-end link has determined transmission performance end to end in network application.Therefore the estimation (measurement) of path capacity has application prospect quite widely.

Conventional path capacity measurement model is to send detection packet at the source in path end to destination, by time interval of destination record detection packet or the time of advent characteristic quantity come the measuring route capacity.According to the send mode of detection packet, the path capacity algorithm for estimating can be divided into the variable packet length technology (Variable Packet Size, VPS) and detection packet to (Packet Pair, PP) two class technology.Owing to being influenced by two layers of store-and-forward unit, the VPS technology produces a certain amount of error, so to the improvement of the algorithm for estimating of path capacity with optimize and mainly be based on detection packet to (PP) technology.

Accurately, fast, the path capacity Measurement Algorithm of low expense is the composition basis of a lot of network applications.The VPS technology is because the inherent shortcoming on the model can not guarantee the accuracy that path capacity is measured.And the PP technology will obtain path capacity accurately, and key is that link measurement samples collection is carried out statistical analysis.Existing algorithm for estimating can not satisfy demands of applications at aspects such as accuracy, detection time and detection amount expenses.

Summary of the invention

The shortcoming of prior art the object of the present invention is to provide a kind of method of measurement based on the right path capacity of bag in view of the above, is used for solving prior art in the problem of aspects such as accuracy, detection time and detection amount expense.

Reach other relevant purposes for achieving the above object, the invention provides a kind of method of measurement based on the right path capacity of bag, it comprises that to send detection packet by the source end to destination right, and destination calculates according to detection packet and scatter the time window sequence the right time of advent; Try to achieve the distribution time series according to scattering the time window sequence again; At last by scattering the path capacity of time series calculating from the source end to destination.

Preferably, thus can be earlier divide by a certain stroll time window length the distribution time samples that arrives detection packet in the above-mentioned path capacity method of measurement and obtain scattering the time window sequence.

Preferably, can also carry out obtaining scattering time series after the median filter to the distribution time window sequence in the above-mentioned path capacity method of measurement.

Preferably, can also define and give up the factor, and calculate the path capacity from the source end to destination again after the distribution time series in the above-mentioned path capacity method of measurement given up.

Preferably, in the above-mentioned path capacity method of measurement, behind the path capacity that records from the source end to destination, can record the path capacity from destination to the source end again with carrying the ACK technology.

Method of measurement based on the right path capacity of bag of the present invention has adopted the PP technology, does not have the intrinsic error of VPS technology; Adopt new statistics, analysis and computational methods that detection packet is handled the distribution time samples collection that arrives, calculate path capacity, the accuracy height, required time is few, calculates, to measure expense little.

Description of drawings

Fig. 1 is the schematic diagram of PP Detection Techniques.

Fig. 2 flow chart in accordance with a preferred embodiment of the present invention.

Fig. 3 is respectively 1M and 10M(bps to the uplink and downlink path capacity according to a preferred embodiment of the invention) measured value that carries out 50 actual measurements.

Fig. 4 is respectively 15M and 30M(bps to the uplink and downlink path capacity according to a preferred embodiment of the invention) measured value that carries out 50 actual measurements.

Embodiment

Below by specific instantiation explanation embodiments of the present invention, those skilled in the art can understand other advantages of the present invention and effect easily by the disclosed content of this specification.The present invention can also be implemented or be used by other different embodiment, and the every details in this specification also can be based on different viewpoints and application, carries out various modifications or change under the spirit of the present invention not deviating from.

Referring to Fig. 1, what Fig. 1 showed is detection packet is measured end-to-end path capacity to (technology) schematic diagram.End to end path definition by from the source end to destination grouping node (equipment that possesses the storage forwarding capability, the router for example) sequence of process.Suppose that the path of process is constant and unique when measuring, so on the path zero background traffic (Cross Traffic, in the time of CT), the maximum data throughput that the IP layer can reach is defined as path capacity.If total jumping figure (Hop) of node is H on the end-to-end path, U _iBe the i hop node at link capacity or the transfer rate of source end on the destination direction, U ₀Be the link capacity of source end, then as shown in Equation (1), up path capacity U is the link capacity minimum value of end node on the destination direction in source on the path.In like manner, we define D _iBe that i jumps link capacity or the transfer rate of destination to the extreme direction of source, D _H+1Be the link capacity of destination, then as shown in Equation (2), downlink path capacity D is the link capacity minimum value of destination node to the extreme direction of source in the path.

$U=\underset{i=0,...H}{\min }{U}_i---\left(1\right)$

$D=\underset{i=1,...H+1}{\min }{D}_i---\left(2\right)$

For the convenience that illustrates, hereinafter only discuss earlier and measure the up path capacity C.

Still referring to Fig. 1, wherein the detection packet of mark 1 and 1 ' expression equal in length is right, mark 2,3,4 represent respectively link nodes (such as, can be store-and-forward unit).As shown in the figure, here suppose that node 2 is identical with the link capacity of node 4, be 3C, the link capacity of node 3 is C, here suppose that each link nodes is that queuing policy is first in first out (FIFO), bag to 1 and 1 ' sent to destination from the source end in the mode of (back-to-back) back-to-back, finish so once transmission and just finish the one-shot measurement sampling.Note bag 1 and 1 ' length be L, right distribution time (Dispersion Time) Δ of definition bag is: the path link nodes receive bag 1 ' last bit begin to receive the time interval that bag 1 last bit ends (here hypothesis be give out a contract for a project 1 earlier ' after give out a contract for a project 1), i.e. the distribution time is bag 1 and bag 1 ' poor at the time of reception at certain link nodes place.As shown in Figure 1, bag is to sending in back-to-back mode at the source end, the distribution time Δ=L/3C of Gu Bao when leaving source end node 2, the distribution time of leaving node 3 is Δ=L/C, owing to depend on the node of link capacity minimum in the path in the distribution time of destination node 4, so be Δ=L/C in the distribution time of destination node 4.When total jumping figure of node on the path is H and background traffic when being zero, bag is C to jumping link capacity through i _iLink nodes the time, the bag to the distribution time Δ before this hop node _InWith the distribution time Δ that leaves after this hop node _OutSatisfy formula (3)

Δ _out＝max{Δ _in,L/C _i} （3）

If hypothesis bag is through in the path during every jumping link nodes, background traffic is zero, then wrap to the distribution time at destination be Δ _R:

${\mathrm{\Δ}}_R=\underset{i=0,...H}{\max }\{L/C_i\}=\frac{L}{\underset{i=0,...H}{\min }\left(C_i\right)}=\frac{L}{C}---\left(4\right)$

C is path capacity end to end in the formula (4).In the time of advent that destination can record each detection packet, calculate bag according to this at the distribution time of destination Δ _RTherefore can come the measuring route capacity according to formula (5):

$C=\frac{L}{{\mathrm{\Δ}}_R}---\left(5\right)$

With reference now to Fig. 2,, Fig. 2 is the basic flow sheet according to measuring route capacity method of the present invention.In step 1, sending regular length from the source end is that n the detection packet of L is right, does not consider under the situation of packet loss in the transmission course, gathers n sample set S={ Δ that wraps the distribution time at destination so ₁, Δ ₂... the .. Δ _n, from S set, mark off subclass S _i={ Δ _i, Δ _I+1... the .. Δ _I+l-1, and | S _i|=l, S _iBe defined as the distribution time window, l is defined as and scatters time window length.Can mark off a length from stroll time samples S set so is that l scatters time window sequence D={ S ₁, S ₂... S _i... S _N-l+1, }.

According to preferred embodiment of the present invention, in measuring sampling step 1, L=1024 byte, n=50, l=16.What deserves to be explained is that those skilled in the art can select other suitable value for use as required to parameter L, n, l.Such as, for L, can select random length for use, preferably but not necessarily, value is the MTU(Maximum Transmission Unit MTU in path), i.e. L=1500(unit: byte).But for particular platform, the windows in this preferred embodiment for example, owing to there is high-speed channel, it is 1024 that L can be set.For n, can be the arbitrary integer more than or equal to 1, but in order to improve certainty of measurement, preferably but not necessarily, value is more than or equal to 50.For l, can select arbitrary integer between 1 to n (when value equaled n, expression was not divided sample set S) for use, preferably but not necessarily, value is more than or equal to 16.

In the practical application, there is noise in link, for reducing noise to the influence of measurement result, can introduce the noise filtering step sample set S is handled.In noise filtering step 2, this preferred embodiment is to stroll time window S _iFilter, those skilled in the art can select the filter algorithm of any appropriate according to actual conditions or needs for use.In this preferred embodiment, select the median filter algorithm for use.Median filter algorithm computational process is as follows:

Calculate and scatter time window S _iThe intermediate value Δ _Mid:

Δ _mid＝median S _i＝median{Δ _i,Δ _i+1,......Δ _i+l-1} (6)

With the intermediate value Δ _MidFor datum quantity obtains S set _i',

$S_i^{\′}=\left\{v\right|v\∈S_i,\frac{{\mathrm{\Δ}}_{\mathrm{mid}}}{2}\≤v\≤2{\mathrm{\Δ}}_{\mathrm{mid}}\}---\left(7\right)$

According to stroll time window sequence D and formula (6), can calculate the stroll time window sequence S after the median filter _f={ S ₁', S ₂' ... S _i' ... S _I+l-1'.

In statistical analysis step 3, the distribution time window S that obtains according to noise filtering step 2 _i', adopt following formula (8) to calculate the stroll time :

${\widetilde{\mathrm{\Δ}}}_i=\frac{\underset{\∀v\∈S_i^{\′}}{\mathrm{\Σ}}v}{\left|S_i^{\′}\right|}=E\left[S_i^{\′}\right]---\left(8\right)$

According to scattering time window sequence S _fAnd formula (7), (8) can calculate the stroll time series

$\widetilde{\mathrm{\Δ}}=\{{\widetilde{\mathrm{\Δ}}}_1,{\widetilde{\mathrm{\Δ}}}_2,......{\widetilde{\mathrm{\Δ}}}_{n-l+1}\}.$

By scattering time series

Can calculate the distribution time

Concrete grammar is:

${\widetilde{\mathrm{\Δ}}}_f=\underset{i=0}{\overset{n-l+1}{\mathrm{\Σ}}}{\widetilde{\mathrm{\Δ}}}_i---\left(9\right)$

Again will

Substitution formula (5) is replaced Δ wherein _R, just can be in the hope of path capacity.

But in order further to improve certainty of measurement, this preferred embodiment by

Calculate

Before, it has been carried out further selection optimization, concrete selection mode is as follows:

Right

Resequence ${\widetilde{\mathrm{\&Delta;}}}^{\&prime;}=<{\widetilde{\mathrm{\&Delta;}}}_1^{\&prime;},{\widetilde{\mathrm{\&Delta;}}}_2^{\&prime;},......{\widetilde{\mathrm{\&Delta;}}}_{n-l+1}^{\&prime;}>({\widetilde{\mathrm{\&Delta;}}}_1^{\&prime;}\&le;{\widetilde{\mathrm{\&Delta;}}}_2^{\&prime;}\&le;{\widetilde{\mathrm{\&Delta;}}}_3^{\&prime;}......{\widetilde{\mathrm{\&Delta;}}}_{n-l+1}^{\&prime;}),$ Here hypothesis is scattered time series Normal Distribution is so use sample average to estimate population mean.The factor is given up in definition Sequence length is h=n-l+1=35, in sequence

Two ends give up, give up length and be

Distribution time series after giving up so

For:

${\widetilde{\mathrm{\&Delta;}}}^{\&prime;\&prime;}=<{\widetilde{\mathrm{\&Delta;}}}_5^{\&prime;},{\widetilde{\mathrm{\&Delta;}}}_6^{\&prime;},......{\widetilde{\mathrm{\&Delta;}}}_{31}^{\&prime;}>$

The time estimated value of then finally taking a walk is

${{\widetilde{\mathrm{\&Delta;}}}_f}^{\&prime;}=\underset{i=5}{\overset{31}{\mathrm{\&Sigma;}}}{\widetilde{\mathrm{\&Delta;}}}_i^{\&prime;}---\left(10\right)$

Again will at last

Substitution formula (5) can must path capacity final estimated value, i.e. formula (11):

$C=\frac{L}{{{\widetilde{\mathrm{\&Delta;}}}_f}^{\&prime;}}---\left(11\right)$

Those skilled in the art can also do other variation to this preferred embodiment, such as, in the not high occasion of required precision, can omit noise filtering step 2; Perhaps in noise filtering rapid 2, select other noise filtering mode for use; Not right in statistical analysis step 3 Select direct substitution formula (5) to calculate path capacity; Also can be right

Adopt other selection mode; Perhaps also adopt the selection mode as this preferred embodiment, but select other numerical value etc. for use to giving up the factor.All these variations all do not depart from design of the present invention.

After recording the up path capacity with above-described embodiment, can also measure the downlink path capacity with identical principle, step.Preferably, can utilize and carry ACK technology one-shot measurement and go out the up-downgoing path capacity.

Carrying ACK(Ackownlegement namely replys) technology refers to that when destination received the detection packet of source end, the packet situation identical with length sent to the source end as the ACK response packet, this packet carries corresponding ACK sequence number as answer logo.Destination receive a detection packet to the time, send response packet to wrap right mode to the source end.When sending the ACK response packet, need the source end up path path capacity value that destination estimates is write ACK response packet load field, send to the source end together.The source end estimates the downlink path path capacity according to the response packet that receives to utilizing identical principle and measuring algorithm for estimating.Utilize above-mentioned measurement algorithm for estimating and carry the ACK technology, just can estimate up-downgoing path path capacity by one-shot measurement.

What Fig. 3 showed is to utilize the preferred embodiment of the present invention uplink and downlink path capacity to be respectively the end value of 50 gained of path measurements of 1M and 10M, as can be seen from Figure 3, this preferred embodiment is that accuracy is fine as a result for the path measurements of 1M, 10M to path capacity.

What Fig. 4 showed is to utilize the preferred embodiment of the present invention uplink and downlink path capacity to be respectively the end value of 50 gained of path measurements of 15M and 30M, as can be seen from Figure 4, the accuracy of a path measurements that preferential path capacity of embodiment is 15M and 30M is also better.

In sum, the invention provides a kind of method of new measuring route capacity, the accuracy height, required time is few, and the expense of calculating, measuring is little.So the present invention has effectively overcome various shortcoming of the prior art and the tool high industrial utilization.

Above-described embodiment is illustrative principle of the present invention and effect thereof only, but not is used for restriction the present invention.Any person skilled in the art scholar all can be under spirit of the present invention and category, and above-described embodiment is modified or changed.Therefore, those skilled in the art must be contained by claim of the present invention not breaking away from all equivalence modifications of finishing under disclosed spirit and the technological thought or changing.

Claims (10)

1. the method for measurement based on the right path capacity of bag is characterized in that, comprises the steps:

Steps A, it is right to receive first detection packet that sends from the source end by destination, according to described first detection packet the distribution time samples collection that arrives is calculated and scatters the time window sequence;

Step B filters described distribution time window sequence and obtains scattering time series;

Step C calculates the distribution time according to described distribution time series, again according to scattering the path capacity of Time Calculation from described source end to described destination.

2. path capacity method of measurement according to claim 1, it is characterized in that, in steps A, described destination is according to scattering time window length, obtain described distribution time window sequence after described first detection packet divided the distribution time samples collection that arrives, described distribution time window length is more than or equal to 16 and smaller or equal to the arbitrary integer of described distribution time samples collection length.

3. path capacity method of measurement according to claim 1 is characterized in that, is calculated by described distribution time series and scatters time method:

Note is scattered time series

Wherein l is described distribution time window length, and n is the distribution time samples collection length of described detection packet to arriving, and then scatters the time

${\widetilde{\mathrm{\&Delta;}}}_f=\underset{i=0}{\overset{n-l+1}{\mathrm{\&Sigma;}}}{\widetilde{\mathrm{\&Delta;}}}_i.$

4. path capacity method of measurement according to claim 1 is characterized in that, described first detection packet centering, and detection packet length value is the MTU in path.

5. path capacity method of measurement according to claim 1 is characterized in that, what described distribution time window sequence was filtered employing is the median filter mode, and described median filter mode comprises:

Remember that each scatters time window S in the described distribution time window sequence _iIn each element be Δ _i, S _iIntermediate value be Δ _Mid, described destination is l according to scattering time window length, then:

Δ _mid＝medianS _i＝median{Δ ₁,Δ ₂,Δ ₃.......Δ _l}

With the intermediate value Δ _MidFor benchmark filters described distribution time window S _iObtain S _i',

$S_i^{\&prime;}=\left\{v\right|v\&Element;S_i^{\&prime;},\frac{{\mathrm{\&Delta;}}_{\mathrm{mid}}}{2}\&le;v\&le;2{\mathrm{\&Delta;}}_{\mathrm{mid}}\}$

Stroll time window sequence after median filter is S _f:

S _f＝{S′ ₁,S′ ₂,S′ ₃.....S′ _i+l-1}

6. path capacity method of measurement according to claim 1 is characterized in that, among the step C, before described distribution time series calculating path capacity, also comprises described distribution time series is selected.

7. path capacity method of measurement according to claim 5 is characterized in that, the mode that described distribution time series is selected is:

To described distribution time series

Resequence and obtain

Factor-alpha is given up in definition,

Sequence length is h=n-l+1, in sequence

Two ends give up, give up length and be

The final distribution time

For:

8. path capacity method of measurement according to claim 6 is characterized in that, the described factor of giving up is smaller or equal to 0.1.

9. according to the described path capacity method of measurement of any one claim of claim 1 to 8, it is characterized in that, also comprise the path capacity of measurement from described destination to described source end, the path capacity of wherein measuring from described destination to described source end comprises the steps:

Step D1, it is right to receive first detection packet that sends from the source end by destination, according to described first detection packet the distribution time samples collection that arrives is calculated and scatters the time window sequence;

Step D2 filters described distribution time window sequence and obtains scattering time series;

Step D3 is according to the path capacity of described distribution time series calculating from described source end to described destination.

10. path capacity method of measurement according to claim 9 is characterized in that, described second batch of detection packet comprises the path capacity information from described source end to described destination to being the ACK response packet in the described ACK response packet.

Images